The amount of traffic that passes through a network fluctuates and a node through which no packets pass will sometimes appear, depending on time zones. There currently exists a node which may operate in a sleep mode and which realizes power saving by operating only the minimum necessary functions and minimizing its power consumption in nodes through which no packets pass.
Among methods for controlling the sleep mode of a node there is one in which a server for collectively managing all of the traffic status in a network is provided and the collective management server instructs a specific node to shift and return to a sleep mode according to the traffic status.
A device capable of determining whether a certain device is in a sleep mode or its power is switched off by transmitting a packet to a network device that may be in a sleep mode from another device and monitoring its response packet is also known.
In a method using a conventional collective management server, it is necessary to dispose in a network a server having a function to collect traffic information, manage it, and make determinations on the basis of it. Such a method may not cope with network so large that it exceeds the process capability of a server. With such a method, it becomes necessary to modify its settings every time the topology of a network or the like is modified and thus management costs become high.
Even if a packet does not arrive for a short time when a node shifts to a sleep mode via a determination by the node that is based only on whether a packet has actually been input to it, sometimes the node shifts to a sleep mode and it becomes necessary to promptly awake (return from a sleep mode) when a packet arrives immediately after the node has begun sleeping. Thus, when a node shifts to a sleep mode via a determination by the node that is based only on whether a packets has actually been input to it, its mode is frequently switched over.
There is also a method of return from a sleep mode in which a node returns when a packet reaches a sleep node or its adjacent node enables a sleep node to be returned to when a packet reaches a node immediately before a sleep node. However, in this case, packet transfer is delayed by an amount of time necessary for it to return.
Patent document 1: Japanese Laid-open Patent Publication No. 2000-261515